I would to recommend chunking in vector search typesense, so when user search by vector_search the typesense can scoring average simillarity from each document (the document have some chunking). Or this feature has been add?